Analysis of CT features of lepidic predominant subtype and other pathological subtypes in early-stage invasive lung adenocarcinoma appearing as ground-glass nodule / 中华放射学杂志

ABSTRACT

Objective:To investigate the CT features of lepidic predominant adenocarcinoma (LPA) and other pathological subtypes in early-stage invasive pulmonary adenocarcinoma appearing as ground glass nodule (GGN); and to provide imaging-derived information for the clinical management of GGN.Methods:The clinical and CT data of patients with early-stage invasive pulmonary adenocarcinoma in the First Medical Center of PLA General Hospital from January to December 2019 were retrospectively reviewed. All patients presented with pure GGNs or mixed GGNs with a consolidation-to-tumor ratio (CTR)<0.5, with the pathological results confirmed by surgery. GGNs were divided into LPA and non-LPA (n-LPA) groups according to pathological subtypes. Univariate analysis was used to compare the clinical data and CT characteristics between the two groups. The multivariate analysis was performed for the indicators with statistically significant differences and a multivariate model was generated using the reverse elimination method. The area under the ROC curve (AUC) was used to evaluate the discriminatory power of this model for differentiation of LPA from n-LPA.Results:A total of 630 GGNs from 589 patients were analyzed, with 367 GGNs in LPA group and 263 GGNs in n-LPA group. In univariate analysis, the diameter [(14±5) mm], CT value [(-566±98) HU], and CTR [13.9% (0, 27.3%)] in the LPA group were significantly smaller than those in the n-LPA group [(15±5) mm, (-499±111) HU, 27.8%(7.7%, 40%)], respectively, P<0.05]. The frequency of mGGN, deep lobulation sign, burrs, vascular changes, bronchial changes, and clear tumor-lung interface were significantly higher in the n-LPA group than those in the LPA group ( P<0.05). Multivariate analysis results showed that mean CT values, CTR, deep lobulation sign, burr, vascular changes, and bronchial changes were independent predictors for predicting n-LPA ( P<0.05), which were included in the logistic model. Using the optimal cutoff value of 3.958, the logistic regression model for differentiate LPA from n-LPA had a sensitivity of 76.4%, a specificity of 78.7%, and an area under the curve of 0.840. Conclusion:The CT features are helpful for differentiating lepidic predominant subtype from other subtypes in early-stage invasive pulmonary adenocarcinoma presenting as a GGN.